Current benchmarks for stream processing systems or SPS cannot hold up with the rise of modern

hardware like FPGAs.

We present SkyShar, a benchmark based on real aircraft tracking data for stream processing

systems utilizing modern hardware.

To track aircraft in high-density flight corridors or over vast land masses and oceans, reliance

on ground-based radar alone is insufficient.

Aircrafts continuously broadcast ADS-B messages containing their identification and position.

ADS-B messages are received by control towers or other official receivers and are utilized

for air traffic management.

These messages are unencrypted and can be received by anyone willing to build their

own software-defined radio.

Broadcasting this data requires immense computing power and plays a crucial part in air traffic

safety.

Open source networks such as OpenSky offer a vast web of receivers globally.

They handle the collection of ADS-B messages and provide an API to download historic data

or fetch real-time data.

Leveraging this publicly available data, we developed a new benchmark.

We formulated 14 queries rooted in real-world applications that utilize not only ADS-B messages

as streaming data, but also incorporate other publicly sourced relational data.

Our benchmark includes simple filter and projection queries designed to run entirely on your FPGA

or other contemporary hardware platforms.

Additionally, we provide more complex queries involving static joins, window functions,

and aggregations designed to test and challenge the capabilities of your system.

Combined with our custom benchmarking tool to measure throughput and end-to-end latency,

we offer a comprehensive package for thoroughly testing and benchmarking your system.

For more information, follow us on skyshark.org.